I'm pretty sure that the difference you have from other DAC is probably due to the output stage as RT6863 is not known (exept for the khadas tone board) and hope it does not include any fancy feature to make the sound pleasant.
(Where to get a datasheet of this RT6863 )
Have you made any ABX test in XLR with another sabre DAC using
2L bench same flavor (24bit352.8kHz and DSD256) ?
Yes with oppo 205 recently. SGD1 is pleasing, dynamic, very good. I listened my DSD files with the very first designs of native DSD playback from cirrus logic, ESS, Texas, AK, most of the time balanced output. Sometime in RCA. Sometime some RCA outputs were better than the XLR and it turned out that it wasn't true balanced output stage. I can say I know my files. DSD playback and specifically DSD is rendering differently here than any Sabre (9016, 9018, 9028, 9038 and their variants)-based design I tried. I like it, but it's very close in the rendering from PCM sometimes. And the 2L test is a classic in that perspective. I read Ben statements with pleasure, but there is still this unusual feeling. Let's give to the SGD1 it's uniqueness of rendering. Going beyond this, the amount of details that is displayed is exceptional.
Again I spent some time listening focusing on the DSD files...the rendering is much different that the oppo, and the matrix I tested also, but some time ago for the matrix.
By the way Ben I came across another bug tonight : From off to On, i used directy a Raspberry runeaudio as a source connected to USB-C. PCM played fine, but no DSD could play... I thought "not again", and I thought maybe you pushed another OTA upgrade for the SGD1 that changed something...Spent some time runing around...switched from RUneaudio to volumio, same problem. Then I came closer to switch from USB-C to USB-B and plug my laptop as source to test with foobar...and it worked suddenly. Even if I was connected to USB-C to the raspberry, the moment I switched to USB-B on the knob, the USB-B light activated and the DSD file was playing from the raspberry....and I could hear the sound from the DAC, even if the Raspberry was connected to the USB-C input. I changed again to USB-C and then both DSD and PCM could play.
There is a inconsitency in input selection and maybe the XMOS input is rebooted everytime you switch from USB-C to USB-B and vice versa...
But Ben, Direct DSD was lagging big time after that, when played from a raspberry Pi 3B (work no problem on other DAC with XMOS U208 entry, who are native DSD from end to end). No problems with DoP mode, until DSD256...it should be the contrary, DoP put a load on the CPU to chunk the DSD signal in PCM bytearrays. And theoretically the XMOS entry, since it can output in I2S (on other things), has capabily to convert, reschedule and check signals...and convert to PCM...
Can you please confirm how the XMOS entry, after your upgrade, handle the DSD, and how it is conveyed to the Sabre twin DACs? Having said that there absolutely NO DSD to PCM conversion whatsoever anywhere in the DAC, I won't open and pin-check the xmos board, I trust Ben. What is exactly the CPLD (or other) rationale for splitting the signal to the two chips in Mono mode? The design has to be well thought and engineered to handle DSD stream from one end to the others when using two chips (it took Audio GD some time and iterations...).
I have pleasure listening to this DAC nonetheless.
Thank you for this really helpful review.
Regarding the DSD playback: from my knowledge lowering the hardware volume on ESS dacs makes DSD playback impossible.
I lowered the volume on my Tone Board (using alsamixer) and when I play DSD
- as DoP there is no sound + bonus the DAC is blocked, it won't play PCM also until a reboot.
- as direct-DSD it makes an annoying fisss sound, If you try that, lower the volume to the min on the amp or better yet use some cheap headphones
This could be a test that the DSD streams you send to the dac are not converted to PCM - i think...
I hope after a few more days when your sound impressions will be clearer you will come back here to share them with us.
This is exactly why I don't recommend putting a Volume knob where there is no amplifier, and that I favor steps volume approach (I have an Audio GD head Amp after all) I hope that this volume circuitry is not inducing a PCM processing along the way. Ben will comment how and where volume is managed. For DSD the Sabre DAC chips have ways to handle digital gain just before the I/V conversion if I remember well. But Volume has to be handled only if you need it, in amplifiers only in my view. Fortunately the SGD1 in his latest firmware memorizes the volume setting.
Fundamental knowledge on DSP.
The only chip in this product that supports general purposed processing is an XMOS XU208, which is underpowered for most DSP calculations. The upper limitation of that chip is to do some simple volume calculations.
Even much more powerful FPGA chips in ADI-2 Pro/DAC has quite a lot of DSP limitations. RME has to shut down DSP features in various cases (high res music, or DSD music).
XMOS has capabilities to convert signal (DSD to PCM normally...), and the product uses a dual mono 9038Q2M design wich requires to distribute perfectly the bits to each chip with perfect timing...I don't see how this is possible without a FGPA/CPLD/programable DSP somewhere, so there ought to be a logic device capable of calculation. Ben, how do you do it?